Microsoft has just offered new tools to detect but also to prevent creating “deepfakes“, these video or photo montages which manipulate the truth and are particularly worrying as the US presidential election approaches.
These assemblies consist for example of “paste“the head of a person on the body of another in a video in order, in particular, to make him say things that he has never said. They can transform the meaning of an event or invent all of them coins, to spread false information or damage the reputation of public figures.
The computer giant announced Monday it has designed a “Video Authenticator“, which analyzes photos and videos to determine if they have been manipulated with artificial intelligence (AI) technologies.
This tool assigns a “trust score“to the content reviewed.”In the case of a video, it can provide a percentage (indicating the probability of manipulation), in real time during playback, frame by frame“, explains the statement.
It detects the mistreatment of content by spotting badly blurred boundaries, inconsistent blurs or colors invisible to the naked eye.
Protect the US Elections
The tech giants, social networks and AI specialists, are redoubling their efforts two months away from the US presidential election to counter the influence campaigns carried out from abroad but also within the country, to sow division and manipulate voters.
In 2016 the US presidential election and the referendum on Brexit in the United Kingdom were marked by vast and powerful operations of this kind, orchestrated from Russia.
According to a study conducted at Princeton University and funded by Microsoft, out of a hundred foreign campaigns that targeted thirty countries between 2013 and 2019, 93% involved the creation of content created from scratch and 74% distorted verifiable facts.
“Methods of producing artificial media will become more and more sophisticatedrecalls the Redmond (near Seattle, West) company, which is in line to buy the TikTok lightweight video sharing app.
–
